Buzzfindr: Automating the detection of feeding buzzes in bat echolocation recordings
Bibliographic record
Abstract
Quantification of bat communities and habitat heavily rely on non-invasive acoustic bat surveys the scope of which has greatly amplified with advances in remote monitoring technologies. Despite the unprecedented amount of acoustic data being collected, analysis of these data is often limited to simple species classification which provides little information on habitat function. Feeding buzzes, the rapid sequences of echolocation pulses emitted by bats during the terminal phase of prey capture, have historically been used to evaluate foraging habitat quality. Automated identification of feeding buzzes in recordings could benefit conservation by helping identify critical foraging habitat. I tested if detection of feeding buzzes in recordings could be automated with bat recordings from Ontario, Canada. Data were obtained using three different recording devices. The signal detection method involved sequentially scanning narrow frequency bands with the "Bioacoustics" R package signal detection algorithm, and extracting temporal and signal strength parameters from detections. Buzzes were best characterized by the standard deviation of the time between consecutive pulses, the average pulse duration, and the average pulse signal-to-noise ratio. Classification accuracy was highest with artificial neural networks and random forest algorithms. I compared each model's receiver operating characteristic curves and random forest provided better control over the false-positive rate so it was retained as the final model. When tested on a new dataset, buzzfindr's overall accuracy was 93.4% (95% CI: 91.5%- 94.9%). Overall accuracy was not affected by recording device type or species frequency group. Automated detection of feeding buzzes will facilitate their integration in the analytical workflow of acoustic bat studies to improve inferences on habitat use and quality.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.003 | 0.007 |
| Meta-epidemiology (narrow) | 0.002 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.004 | 0.007 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".